JP2003085480A - Method and device for detecting vehicle loaded with specific object - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device for detecting a vehicle loaded with specific objects to exactly judge the vehicle loaded with the specific objects by exactly recognizing characters, for example, 'danger', 'poison', 'toxic', for indicating the specific objects loaded in a traveling vehicle. SOLUTION: The vehicle loaded with the specific object is detected by generating a high-frequency image by first applying the minimum value filter and then applying the maximum value filter to a picked up image including a plate provided on the vehicle on which the characters to indicate the vehicle loaded with the specific object are described, generating an image to definitize character parts of the picked up image by taking AND between an image by inverting a negative image created by subtracting the picked up image from the image generated by applying the maximum value to the picked up image and the high-frequency image, extracting character areas from the image and recognizing the characters indicating the vehicle loaded with the specific objects.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for recognizing a character written on a plate or the like provided on a traveling vehicle to detect a vehicle carrying a specific object,
In particular, the present invention relates to a method and apparatus for detecting a vehicle carrying a specific object by accurately detecting characters such as “danger,” “poison,” and “play” that indicate a vehicle carrying a specific object. is there.

[0002]

2. Description of the Related Art Vehicles carrying specific substances such as oils and chemicals, poisonous substances, and specific substances such as powerful drugs must be sized by law to show that they are vehicles that carry these specific substances. Write the letters such as "danger,""poison," and "play" in white or yellow on the black-painted plate, and attach it around the number plates of the front and rear bumpers of the car as shown in Fig. 11. Is obligatory. Such vehicles carrying a specific object are prohibited from entering, for example, in a long tunnel, etc., and only a vehicle carrying a specific object that has been permitted is allowed to pass in a specific place. However, some vehicles loaded with a specific object inadvertently enter such a prohibited area, and a device for determining a vehicle loaded with a specific object is required to detect such a vehicle. .

To specify such a vehicle carrying a specific object, for example, "danger" indicating the vehicle carrying the specific object described above,
It is the simplest to recognize characters such as "poison" and "drama", but as a device for detecting characters on a plate attached to a vehicle in this way, for example, Japanese Patent Laid-Open No.
In Japanese Patent No. 193266, in order to read a license plate of a vehicle at a tollgate on a highway, an imaging device for reading the license plate and a device for detecting the entrance of the vehicle are installed in each lane, and if the entrance of the vehicle is detected, There is one in which an image of the license plate is picked up and read by an image pickup device.

[0004]

However, the device disclosed in Japanese Unexamined Patent Publication No. 2-193266 has a problem that the speed of a vehicle becomes lower than a certain speed, such as a tollgate on a highway, and that each vehicle is prevented from entering. It is a license plate recognition device in a place where it can be detected separately from each other, such as slowing down the speed of a passing vehicle such as the entrance of a tunnel described above, or detecting each vehicle traveling in multiple lanes. It cannot be used in a place that is difficult to recognize separately.

Further, since the device disclosed in Japanese Patent Laid-Open No. 2-193266 is a license plate recognition device whose installation location is uniquely determined, the above-mentioned "danger",
If there is a certain degree of freedom in the installation location in the vehicle, such as a plate for "poison" or "play", such a license plate recognition device cannot accurately recognize the license plate.

[0006] Unlike the license plate, the plate on which characters such as "danger,""poison," and "drama" that indicate such a vehicle loaded with a specific object are attached is vulnerable to dirt and scratches. In addition to the occurrence of, the contrast of characters may be reduced and recognition may be difficult.

Therefore, in the present invention, the characters indicating the vehicle carrying the specific object, such as "danger,""poison," and "play" in the running vehicle are accurately recognized, and the vehicle carrying the specific object is accurately determined. An object of the present invention is to provide a vehicle for detecting a specific object loaded.

[0008]

In order to solve the above problems, a first aspect of the present invention is a method invention, wherein a traveling vehicle is imaged by an image pickup device, and the image pickup result is subjected to image processing to obtain a vehicle carrying a specific object. A method for detecting a vehicle carrying a specific object by reading the character shown to detect the vehicle carrying a specific object,
An image captured by the image capturing device including a character indicating a vehicle loaded with a specific object in a traveling vehicle is subjected to a minimum value filter having a minimum luminance in a pixel of a predetermined size as a central pixel luminance, and then a maximum in a pixel of a predetermined size. A high-frequency image is generated by applying a maximum-value filter having brightness as a central pixel brightness, and a negative image created by subtracting the captured image from the image generated by applying the maximum-value filter to the captured image and the high-frequency image. An image is created by clarifying the character part in the picked-up image by taking the logical product of and, and the character area is extracted from the image and the character indicating the vehicle carrying the specific object is recognized to detect the vehicle carrying the specific object. It is characterized by doing so.

According to a fifth aspect of the present invention, which is a device invention for carrying out the method invention, an image pickup device for picking up an image of a traveling vehicle is provided.
An image processing device that recognizes a character indicating a vehicle loaded with a specific object from an image captured by the imaging device, detects a character indicating the vehicle loaded with the specific object, and detects a vehicle loaded with the specific object A device, wherein the image processing device is
A high-frequency image obtained by applying a minimum value filter having the minimum luminance of pixels of a predetermined size as the central pixel luminance and a maximum value filter having the maximum luminance of pixels of the predetermined size as the central pixel luminance to the image captured by the imaging device. A high-frequency image generating means, a negative image generating means for generating an inverted image of a negative image created by subtracting the captured image from an image generated by applying a maximum value filter to the captured image by the imaging device, and the negative A means for extracting a character area by taking a logical product of the image generating means and the generated image of the high frequency image generating means, and a character recognizing means for recognizing a character indicating the specific object loading vehicle from the area extracted by the character area extracting means. And a character indicating a vehicle carrying a specific object by clarifying a character portion in a captured image by a logical product of the images generated by the negative image generating means and the high-frequency image generating means. Characterized in that to detect the specific stacking vehicle by recognizing.

As described above, a high-frequency image is generated by using the minimum value filter and the maximum value filter for the captured image, and at the same time, the captured image is subtracted from the resulting image by applying the maximum value filter to the captured image. By taking the logical product of the inverted image and the high-frequency image, an image in which the thickened characters and noise in the captured image are removed is obtained, and the contrast of the plate showing the vehicle carrying the specific object is deteriorated due to contamination. It also enables accurate character recognition.

The extraction of the character area is defined in claims 2 and 6.
As described above, the extraction of the character region is performed by extracting a block in which white pixels are continuous in the image and setting a circumscribed rectangle for each block. The character area extracting means extracts a block in which white pixels are continuous in an image obtained by a logical product of the negative image generating means and the generated image of the high frequency image generating means, and circumscribes each extracted block. It is characterized by being configured by means for setting a rectangle.

From the image in which the picked-up image is clarified in this way, a portion in which white pixels are continuous blocks is extracted,
By setting the circumscribed rectangle of this extracted block,
It becomes possible to extract character areas easily and at high speed.
By checking the block corresponding to the size of the character indicating the vehicle loaded with the specific object from this character area, the recognition of these characters becomes easy.

Further, in the extraction of the character area, when the character indicating the vehicle carrying the specific object is not made of a single block but two blocks are combined, the extraction is made according to claim 3 or 7. As described above, when the plurality of circumscribed rectangles are overlapped with each other, or when there is a specific positional relationship in one of the upper, lower, left, and right directions, the extraction of the character region is a rectangle obtained by merging the plurality of circumscribed rectangles. It is characterized by doing in. The character area extracting means has a plurality of overlapping circumscribed rectangles,
Or if there is a specific positional relationship in one direction of up, down, left and right,
It is characterized in that it has means for merging the plurality of circumscribing rectangles, and extracts a character area from the rectangles merged by the merging means.

When a plurality of circumscribing rectangles set in this way are overlapped with each other or have a specific positional relationship in one direction of up, down, left and right, extraction of a character area is performed by merging the plurality of circumscribing rectangles. By doing so, it is possible to accurately recognize the character indicating the vehicle loaded with the specific object.

In the method and apparatus for detecting a vehicle loaded with a specific object, as described in claims 4 and 8, when the characters indicating the vehicle loaded with the specific object are "danger", "poison", and "drama". Especially effective for.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be exemplarily described in detail below with reference to the drawings. However, the dimensions, materials, shapes, relative positions, and the like of the constituent parts described in this embodiment are not intended to limit the scope of the present invention thereto, unless otherwise specified, and are merely It is only an example.

FIG. 1 is a schematic block diagram of an apparatus showing an embodiment of the present invention, FIGS. 2 to 8 are flow charts of a method for detecting a vehicle carrying a specific object of the present invention, and FIG. 9 is a specific object of the present invention. FIG. 10 is a diagram for explaining a processing state of an image in the loaded vehicle detection method, and FIG. 10 is a diagram for explaining a character area cutout process for recognizing a character indicating a vehicle loaded with a specific object.
It should be noted that in the following description, the characters indicating the vehicle loaded with the specific object will be described as an example of "danger", "poison", "drama", etc. However, it is obvious that the present invention is not limited to these characters. .

In FIG. 1, 1 is a road, 2 is a passing vehicle,
Reference numeral 3 denotes a vehicle detection image pickup device 4, a gantry for installing a light source 5 such as a vehicle irradiation strobe on the road, 6 denotes an optimum position on the road 1 for picking up an image of a license plate of the vehicle 2, 7 Is an analog / digital conversion circuit for analog / digital converting the signal from the image pickup device 4, 8 is an image smoothing unit including a commonly used smoothing filter for removing and suppressing noise in the image, Reference numeral 9 denotes a high-frequency image generation unit that includes a minimum value filter and a maximum value filter, which will be described later, and that extracts a frequency of a certain frequency or more from the captured image. 10 denotes a captured image (input image) from the image generated by the maximum value filter. A negative image generation unit that reverses the obtained image, 11 is a logical product generation unit that takes the logical product of the images generated by the high frequency image generation unit 9 and the negative image generation unit 10, and 12 is a sentence The area extraction unit extracts a block in which white pixels are continuous in the image, and a labeling unit 13 that sets a circumscribing rectangle for each block and a plurality of the set circumscribing rectangles overlap each other or vertically. When there is a specific positional relationship in one of the left and right directions, the merge unit 14 that merges the plurality of circumscribed rectangles is included. Reference numeral 15 is a character recognition unit for recognizing a character indicating a vehicle carrying a specific object such as “danger”, “poison”, and “play” or a character on a license plate from a region extracted by the character region extraction unit 12 by a known character recognition unit. , 16 are recognized letters and license plates,
Specific-object-carrying vehicle determination unit for specifying a specific-object-carrying vehicle, 17
Is a light source emission control unit that controls the light source 5 such as a strobe to emit light, and 18 is a timing (position) optimum for detecting the position of the vehicle from the captured image and reading a character or license plate indicating a vehicle carrying a specific object. It is a vehicle detection unit that sends a signal to the light source light emission control unit 17 to cause the light source to emit light when it becomes low.

In the present invention, for example, a plurality of image pickup devices 4 and a light source 5 such as a strobe are provided on a gantry 3 provided across the road 1, and the plurality of image pickup devices 4 cover the entire width of the road 1. A vehicle traveling at any position on the road 1 can be captured by capturing an image. Then, when the vehicle 2 is detected by the vehicle detection unit 18, when the vehicle 2 comes to the image pickup position 6 on the road 1 (the image pickup position 6 in the drawing is a temporary frame shown for convenience of explanation), The light source 5 such as a strobe is made to emit light to recognize the characters on the license plate of the vehicle 2 or the characters such as “danger,” “poison,” and “play” posted on the vehicle loaded with the specific object. Although the image pickup device 4 and the light source 5 such as a strobe are set on the gantry 3, it is obvious that the image pickup device 4 and the light source 5 such as a strobe may be installed on a columnar shape standing on the roadside.

The vehicle detection unit 18 for discriminating the traveling vehicle has a memory for storing a plurality of images as disclosed in, for example, Japanese Patent Laid-Open No. 4-111079, and compares the brightness at the same address of the memory. The method of recognizing the changed portion as a moving body, the method of detecting the movement of one point on the vehicle as a vector as in Japanese Patent Application Laid-Open No. 7-50769, and the method of detecting the movement of one point on the vehicle as a vector, are followed by each other. A method of sequentially specifying the position of a moving body from imaged data using a template matching method, IEICE Transactions (D-II) Vol. J80-D-II, No.
5 (1997) "Detection of Moving Object Under Complex Fluctuating Background", and Mitsubishi Heavy Industries Vol. 36, No. 2 (19
83) As described in "Development of moving body monitoring system using moving image processing technology", etc., only the moving body area that moves in a uniform linear motion is detected by detecting the optical flow from the imaged data and using the voting method. The light source emission control unit detects the timing when the vehicle comes to an optimum position for detecting a character indicating a vehicle loaded with a specific object on the vehicle by using the method of A signal is sent to 17 to turn on the light source 5 for recognition of these characters.

First, the outline of a specific object loading vehicle detection method and apparatus of the present invention will be described with reference to FIGS. 1, 2, and 9. When the vehicle loaded with the specific object approaches, the vehicle detection unit 18 detects this and when the vehicle arrives at the position 6 on the road in FIG. 1, it sends a signal to the light source emission control unit 17 to cause the light source 5 to emit light. The imaging device 4 images the front surface 110 of the vehicle traveling on the road 1 as shown in FIG. Then, in step S1 in FIG. 2, the signal is analog / digital converted by the analog / digital conversion circuit 7, and steps S2 and S are performed.
In 3, the image smoothing unit 8 thins out the image to 1/2 × 1/2 in order to increase the processing speed, and further becomes a signal in which noise and the like are removed by a smoothing filter of about 3 × 3 pixels. An image as shown in FIG. 9A is obtained. It should be noted that thinning out the image to 1/2 × 1/2 here means that when a vehicle is imaged at a predetermined position 6 on the road, for example, “danger”,
This is because it is possible to take a relatively large image of a character indicating a vehicle carrying a specific object such as “poison” or “drama”.

Then, in step S4, the high-frequency image generation unit 9 uses the minimum value filter having the minimum luminance of the constant pixels as the central pixel luminance and the maximum value filter having the maximum luminance of the constant pixels as the central pixel luminance. A binary image of a high-frequency image as shown in FIG. 9B is generated, and subsequently, in step S5, the maximum value filter in the negative image generation unit 10 is used to thicken the character image, and the input image is converted from the image. The negative image created by subtraction is inverted to generate a binarized image as shown in FIG. 9C, and in step S6, the logical product generation unit 11 causes the binary image of the high-frequency image and the negative image. Is inverted to obtain a logical product of the binarized images to clarify the character portion in the captured image, and an image as shown in FIG. 9D is generated.

Then, in step S7, the character area extracting unit 12
The label merging unit 1 extracts a block in which white pixels are continuous in the image and sets a circumscribed rectangle for each block.
In 4, for example, when the upper stroke and the self portion in “danger” indicating the vehicle loaded with the specific object are labeled separately, the character areas are extracted by merging them. Then, in step S8, the character recognizing unit 15 recognizes the character indicating the vehicle loaded with the specific object, determines whether the vehicle is the vehicle loaded with the specific object, and outputs the determination result of the vehicle loaded with the specific object in step S9.

The above is the outline of the operation of the specific object loading vehicle detecting method and apparatus according to the present invention. Next, the detailed operation of the specific object loading vehicle detecting method and apparatus of the present invention will be described with reference to FIGS. 3 to 8. To do.

First, regarding the high-frequency image generation and the binarized image generation in step S4 in FIG. 2, as shown in FIG. 9A in which the image of the front surface 110 of the vehicle in FIG. In this image, the letters such as "danger" and letters on the license plate are white or yellow, so it is brightly imaged, and since the radiator part on it is bright in places, only that part is white and below the bumper of the vehicle. Is almost darkly imaged. Therefore, the character part and the radiator part above it become a high-frequency image with a bright pattern and a dark pattern with a certain period or more, and the dark part below the bumper basically has no bright or dark pattern. It becomes a low area. Therefore, the high frequency image generation in step S4 is a process for extracting this high frequency image.

In this high frequency image generation processing, first, the minimum value filter is applied in step S41 of FIG. As shown in step S41, the minimum value filter first prepares a 99 × 1 mask which is about 5 dots larger than the character width such as “danger” indicating a vehicle loaded with a specific object. Then, with respect to the image as shown in FIG. 9A obtained by the image smoothing unit 8 of FIG. 1, this mask is sequentially applied from the end, and the dot of the minimum brightness in 99 dots respectively applied is the brightness of the center dot of this mask. And This way,
The image in FIG. 9A has a shape in which a bright portion is embedded in a dark portion around the image, and characters are dark. In this description, the minimum value filter is assumed to be 99 × 1 pixels, but it is obvious that the number of pixels is not limited to this, and this is the same in the description of the maximum value filter below.

Next, in step S42, the maximum value filter is applied to the image subjected to the minimum value filter. Like the minimum value filter, this maximum value filter is 99 × 1 which is about 5 dots larger than the character width such as “danger” indicating a vehicle carrying a specific object.
First, the mask is prepared, and this mask is sequentially applied from the edge of the image to which the minimum value filter has been applied, and the dot having the maximum brightness of 99 dots is the brightness of the central dot of this mask. In this way, the characters darkened by the minimum value filter become almost the original image, but the noise component is removed during this.

Then, in the next step S43, the image subjected to the maximum value filter is subtracted from the input image as shown in FIG. 9A obtained by the image smoothing unit 8 in FIG.
The image as shown in FIG. 9A obtained by the image smoothing unit 8 in FIG. 1 has a threshold value because the character "danger" representing a vehicle carrying a specific object is dirty or the character cannot be accurately drawn if the contrast is poor. Is set low, and the characters are drawn thick. However, when the minimum value filter and the maximum value filter are applied in this way and the image subjected to this filter is subtracted from the original image, the low-frequency component is canceled and the very bright part of the character part is subtracted from the collapsed part. It becomes a very clear shape when burned.

Next, in step S44, the obtained image is binarized by the 5% P tile method, for example. The P tile method is to obtain the ratio of the area of the pattern to be extracted and the area of the background in the image, and use the value that gives the same ratio of the luminance frequency on the luminance histogram as the threshold value. Effective if known. The value of 5% needs to be set to an appropriate value depending on the field of view of the image to be captured, the complexity of the background, etc., but about 5% is appropriate.

Next, regarding the negative image generation and the binary image generation thereof in step S5 in FIG. 2, and the inversion of the image, FIG. 9A obtained by the image smoothing unit 8 of FIG.
As shown in FIG. 4, the negative image generating unit 10 first applies a maximum value filter with about 7 × 7 pixels in step S51 for such an image. Then, since the characters in the image become thick, in the next step S52, the input image is subtracted from the image obtained by applying the maximum value filter. Then, if, for example, 50 out of 256 gradations is binarized with a fixed threshold value, a binarized image with an edge as shown in FIG. 9C is obtained. I'll flip it over. In this description, the maximum value filter is based on 7 × 7 pixels, but it is obvious that the number of pixels is not limited to this. Further, the value of the fixed threshold value 50 for binarization needs to be set to an appropriate value depending on the intensity of the illumination constituting the device, but about 50 is suitable.

Then, as shown in step S6 in FIG. 2, the logical product of the binary image of the high-frequency image and the image obtained by inverting and binarizing the negative image is obtained by the logical product generation unit 11 shown in FIG. , An image as shown in FIG. 9D in which the character portion in the captured image is clarified is generated. Characters can be clarified by multiplying logical products with different binary images in this way, and even in places where it would not be possible without zooming up in the past, it was possible to identify characters without being affected by noise or low contrast. You will never miss a letter that shows a loaded vehicle.

When the characters have been clarified in this way, the character region extraction in step S7 in FIG. 2 is carried out. Labeling of character candidates is first performed in step S71 as shown in FIG. This labeling is performed by observing a part where the white pixels are continuous in the obtained image under various conditions and taking out a part which is judged as one block and assigning a number to the part. It is to clarify whether there is a block. In other words, the size of characters such as "danger,""poison," and "drama" that indicate a vehicle loaded with a specific object is determined by law. It can be understood that the block of character candidates can be roughly classified by this labeling.

Then, in step S72, if the labeled block is a bias or a whit which constitutes a character, a merge is performed to combine it. That is, for example, the letters "danger" and "poison" representing a vehicle loaded with a specific object include, for example, "danger" in (A) includes the block indicated by 100, as shown in FIG. Also, the "poison" in (B) is 10
The “main” block shown in 1 and the “mother” block shown in 102 are located above and below. However, the above-mentioned step S
In the labeling of No. 71, these may be labeled completely separately, and if the characters are recognized as they are, they may not be recognized as "danger" or "poison". Therefore, in this step S72, these are merged by the method shown in FIG.

In FIG. 6, first, step S7 is performed.
Each block labeled from 01 to 704 is registered. Therefore, first, i = 0 is set in order to register all the labels in step S701, and i = i + is set in step S702.
1 is calculated and the first label information is set in step S703. This label information has the following values. Flag [i] = 1 label x start point [i] = label x start point [i] label y start point [i] = label y start point [i] label x end point [i] = label x end point [i] label y end point [i ] = Label y end point [i] label x size [i] = label x size [i] label y size [i] = label y size [i] label area [i] = label area [i] ] Is a flag for determining whether or not this block has been merged. Then, in step S704, it is determined whether i has reached the number of labels, and this cycle is repeated until i <the number of labels.

Then, in the next steps S705 to S710, the cycle is repeated to check all combinations of these labels. Therefore, step S7
In step 05, i = 0 and j = 0 are set, and in step S706, i =
Calculate i + 1. Then, in step S707, it is checked whether or not the flag [i] is used in the merge (when the flag is 0, it is a block already used in the merge), j = i + 1 is calculated in step S708, and step S70 is performed.
In case of the same block in 9 (i = j), and step S7
At 10, it is checked whether the flag [j] is used for the merge, and the process is moved to A or B in each case.

Then, in the next steps S711 to S715, merge processing is performed on the upper part of the character "danger" and the block in the range indicated by 100. In this process, first, in step S711, it is determined whether or not there is an overlapping point between the label [i] and the label [j]. If there is, it is determined in step S712 that the label [i] and the label [j] overlap by 1/2 or more. Make a decision. When it is determined that they overlap, in step S713, the label [i]
And the label [j] are merged, it is determined whether the next value falls within the threshold range. Label area ≤ label area upper limit threshold label circumscribing rectangle x size ≤ label circumscribing rectangle x size upper limit threshold label circumscribing rectangle y size ≤ label circumscribing rectangle y size upper limit threshold And if all of these match, the two blocks are the characters "danger" Step S
Proceed to 714 to merge both labels, and step S715
Then, the flag [j] is set to 0 (that is, the block of j was used for the merge) and the process proceeds to B.

On the other hand, if No in any of steps S711, 712, and 713, the process proceeds to step S716, in which the block "main" indicated by 101 and the block "mother" indicated by 102 in the character "poison" are selected. Perform merge processing. This process is performed by first checking the following items regarding the degree of overlap in the x direction in step S716. Deviation of left end of label [i] and label [j] <threshold value (for example, 10) Deviation of right end of label [i] and label [j] <threshold value (for example, 10) If the overlap in the x direction is within this threshold value, Similarly, in step S717, it is determined whether the distance between the label [i] and the label [j] in the y direction is within a threshold value (for example, 10). If this value is also within the threshold value, in the next step S718, the y-size ratio of the label [i] and the label [j] is checked as follows. Ratio of y size of label [i] and label [j] ≤ threshold value (for example, 0.9) Ratio of y size of label [i] and label [j] ≥ threshold value (for example 1.1) And this value is also within the threshold value If, then the next step S71
When the label [i] and the label [j] are merged in 9,
Check whether the label area and the circumscribed rectangle value are within the following thresholds. Label area ≤ label area upper limit threshold label circumscribing rectangle x size ≤ label circumscribing rectangle x size upper threshold threshold label circumscribing rectangle y size ≥ label circumscribing rectangle y size upper threshold label circumscribing rectangle aspect ratio ≤ circumscribing rectangle aspect ratio lower limit label circumscribing rectangle aspect ratio ≧ circumscribing rectangle aspect ratio upper limit threshold value and if these values are all within the threshold value, it is highly likely that these two blocks are “main” and “mother” that constitute “poison”, and both labels are determined in step S720. Are merged, and the flag [j] is set to 0 (that is, j in step S721).
Used in the merge), and N in any of these steps S716, 717, 718, and 719
The process proceeds to step S722 including the case of o.

Then, in this step S722, it is judged whether or not j has reached the number of labels (J <number of labels), and No.
If so, the process returns to C, the next block is examined, and Y
In the case of es, it is checked in step S723 whether i has reached the number of labels (i <the number of labels), and in the case of No, the process returns to D to check the next block and check the mergeable block. Go on. In the above explanation, merging of the characters "danger" and "poison" was explained,
It is self-evident that in the case of "play", it can be merged in the same manner as in the case of "poison", and by doing so, it is obvious that the same can be done for other characters.

When the merging of step S72 in FIG. 5 is completed in this way, the process proceeds to step S73, and labels including these merged blocks as character area candidates are selected. This process starts with i in step S73.
Is set to 0, i = i + 1 is calculated in step S74, and it is determined in step S75 whether the label area and the size of the label circumscribed rectangle are within the following threshold values, respectively. Label area ≥ Label area lower limit threshold Label area ≤ Label area upper limit threshold Label circumscribed rectangle x size ≥ Label circumscribed rectangle x size Lower threshold threshold label circumscribed rectangle x size ≤ Label circumscribed rectangle x size upper threshold label circumscribed rectangle y size ≥ Label circumscribed rectangle y Size lower limit threshold label circumscribed rectangle y size ≧ label circumscribed rectangle y size upper limit threshold If all these values are within the threshold, the block is registered as a character area candidate in step S76,
The process proceeds to step S77 including the case where these values deviate from the threshold. Then, in step S77, it is determined whether or not all blocks have been processed (i <number of labels). If not completed, the process returns to step S74, and if completed, the process proceeds to step S78 to extract the extracted character region. Remembered.

When the character region is extracted in this way, the character recognition unit 15 and the specific object loaded vehicle determination unit 16 in FIG.
Then, the specific object loaded vehicle determination in step S8 in FIG. 2 is performed according to the flow shown in FIG. That is, first, in step S81, for example, a "danger" detection flag, a "poison" detection flag, and a "drama" detection flag are initialized and set to 0, respectively. And steps S82 and S83
I = 0, i to check all the character areas extracted in
= I + 1 is calculated, the character area stored in step S84 is cut out, and character recognition is performed by a known character recognition method in step S85. And step S8
It is determined in 6, S87 and S88 which character is recognized, the flag corresponding to the character recognized in steps S89, S90 and S91 is set to 1 and the process proceeds to step S92. Also, when there is no character corresponding to the area in which the character recognition is performed, No is determined in step S88 and step S9 is performed.
Go to 2. Then, in step S92, it is determined whether or not i has reached the number of labels in the character recognition area (i <number of labels). If not, the process returns to step S83, and if i, the determination is made in step S93. The result comes.

When the characters "danger", "poison", and "drama" indicating that the vehicle is loaded with the specific object are thus recognized and the vehicle is loaded with the specific object, the flow shown in FIG. A vehicle carrying a specific object is specified by. This is step S
In 95, "danger" indicating that it is a vehicle loaded with a specific object,
It is determined whether or not the "poison" and "drama" plates are present, and if they are present, license plate recognition processing is further performed in step S96 in the same manner as described in the flow of FIG. 7, and in step S97. Vehicle identification processing based on the license plate information is performed, and in step S98, the specific object loaded vehicle determination result is output.

In this way, the characters "danger,""poison," and "drama" indicating that the vehicle is loaded with a specific object are recognized.
Even if the plate on which these characters are written is dirty or scratched and the contrast of the characters themselves is poor, these characters can be accurately recognized and the vehicle carrying a specific object is not missed. In the above description, the characters indicating the vehicle loaded with the specific object are “danger”, “poison”, and
Although the case of "drama" has been described as an example, it is obvious that the present invention can be used to recognize not only these characters but also other characters.

[0043]

As described above, according to the present invention described in claims 1 and 5, a high frequency image is generated by using a minimum value filter and a maximum value filter for a captured image, and at the same time, a maximum value filter is used for the captured image. By taking the logical product of the image and the high-frequency image that is obtained by inverting the negative image generated by subtracting the captured image from the resulting image, an image in which the thickening and noise of the characters in the captured image are removed is obtained, Accurate recognition is possible even if the plate indicating "danger", "poison", "drama", etc., indicating the vehicle loaded with the specific object becomes dirty and the contrast decreases.

According to the present invention described in claims 2 and 6, a portion in which white pixels are continuous blocks is extracted from the image in which the captured image is clarified, and a circumscribed rectangle of the extracted block is set. By doing so, it is possible to easily and quickly extract a character area, and a block corresponding to the character size such as “danger,” “poison,” or “play” that indicates a vehicle loaded with a specific object can be extracted from this character area. Checking makes it easier to recognize these characters.

Further, according to the present invention as set forth in claims 3 and 7, when a plurality of set circumscribing rectangles are overlapped with each other or there is a specific positional relationship in one direction of up, down, left and right, the plurality of the circumscribed rectangles are set. By extracting the character area with the rectangle obtained by merging the circumscribing rectangles, it is possible to accurately recognize the characters such as “danger,” “poison,” and “drama” indicating the vehicle loaded with the specific object.

In the method and apparatus for detecting a vehicle loaded with a specific object according to the present invention, as described in claims 4 and 8, the characters indicating the vehicle loaded with the specific object are "danger", "poison", and "drama".
Is especially effective when

[Brief description of drawings]

FIG. 1 is a schematic block diagram of an apparatus showing an embodiment of the present invention.

FIG. 2 is a schematic flow chart of a specific object loaded vehicle detection method of the present invention.

FIG. 3 is a flowchart of a high frequency image generation method in the specific object loading vehicle detection method of the present invention.

FIG. 4 is a flow chart of a negative image generating method in the specific object loading vehicle detection method of the present invention.

FIG. 5 is a flow chart of a character area extraction method in the specific object loading vehicle detection method of the present invention.

FIG. 6 is a flow chart of label merging in the specific object loaded vehicle detection method of the present invention.

FIG. 7 is a flow chart for determining a vehicle carrying a specific object in the method for detecting a vehicle carrying a specific object according to the present invention.

FIG. 8 is a flow chart for identifying a vehicle carrying a specific object in the method for detecting a vehicle carrying a specific object of the present invention.

FIG. 9 is a diagram for explaining an image processing state in the specific object loading vehicle detection method of the present invention.

FIG. 10 is a diagram for explaining processing for cutting out character areas such as “danger” and “poison”.

FIG. 11 is a view for explaining a mounting position of a plate showing a vehicle carrying a specific object.

[Explanation of symbols]

1 road 2 passing vehicles 3 gantry 4 Vehicle detection imaging device 5 light sources Position on 6 roads 7 Analog / digital conversion circuit 8 Image smoothing unit 9 High-frequency image generator 10 Negative image generator 11 Logical product image generator 12 Character area extractor 13 Labeling section 14 Merge Department 15 Character recognition part 16 Vehicles with specific objects loaded 17 Light source emission control circuit 18 Vehicle detector

Continued front page    F term (reference) 5B029 CC28 EE08                 5B057 BA11 CA08 CA12 CA16 CE06                       CE12 CH09 DA08                 5H180 AA13 CC04 CC24 EE07                 5L096 BA17 DA02 EA35 EA43 FA44                       GA22 HA08 HA13

Claims (8)

[Claims] 1. A method for detecting a vehicle loaded with a specific object, wherein an image of a traveling vehicle is captured by an image capturing device, the imaged result is subjected to image processing to read a character indicating the vehicle loaded with the specific object, and the vehicle loaded with the specific object is detected. Then, the image captured by the image capturing device including the characters indicating the vehicle loaded with the specific object in the traveling vehicle is subjected to the minimum value filter having the minimum luminance of the pixels of the predetermined size as the central pixel luminance, and then the pixel of the predetermined size is obtained. A high-frequency image is generated by applying a maximum value filter having the maximum brightness in the central pixel brightness as the central pixel brightness, and the negative image created by subtracting the captured image from the image generated by applying the maximum value filter to the captured image and the inverted image An image with the character part in the captured image is generated by taking the logical product with the high-frequency image, and the character area is extracted from the image to recognize the character indicating the vehicle carrying the specific object. Specific stacking vehicle detection method is characterized in that to detect the specific stacking vehicle in the. 2. The identification according to claim 1, wherein the extraction of the character region is performed by extracting a block in which white pixels are continuous in the image and setting a circumscribed rectangle for each block. Detection method for loaded vehicles. 3. The extraction of the character area is performed by merging the plurality of circumscribed rectangles when a plurality of the set circumscribed rectangles overlap each other or when there is a specific positional relationship in one of upper, lower, left, and right directions. The method for detecting a vehicle loaded with a specific object according to claim 2, wherein the method is performed in a rectangular shape. 4. The character indicating the vehicle carrying the specific object is
The method for detecting a vehicle loaded with a specific object according to claim 1, wherein the method is "danger", "poison", or "play". 5. An image pickup device for picking up an image of a traveling vehicle, and an image processing device for recognizing a letter indicating a vehicle carrying a specific object from an image picked up by the image pickup device, and detecting a character showing the vehicle carrying a specific object. A specific object loaded vehicle detection device for detecting a vehicle loaded with a specific object, wherein the image processing device has a minimum value filter having a minimum brightness in pixels of a predetermined size in a captured image by the imaging device as a central pixel brightness. A high-frequency image generating means for generating a high-frequency image by applying a maximum-value filter having the maximum brightness in pixels of a predetermined size as the central pixel brightness; and an image generated by applying a maximum-value filter to the image captured by the imaging device. A negative image generating means for generating an inverted image of the negative image created by subtracting the captured image from the negative image generating means, and a logical product of the negative image generating means and the generated images of the high frequency image generating means And a character recognition means for recognizing a character indicating the specific object loading vehicle from the area extracted by the character area extraction means, the negative image generation means and the image generated by the high frequency image generation means A specific object loading vehicle detection device characterized in that a specific object loading vehicle is detected by clarifying a character portion in a captured image by logical product and recognizing a character indicating a specific object loading vehicle. 6. The character area extracting means extracts a block in which white pixels are continuous in an image obtained by a logical product of the negative image generating means and the generated image of the high frequency image generating means, and the extraction is performed. The vehicle for detecting a specific object loading vehicle according to claim 5, characterized in that it comprises means for setting a circumscribed rectangle for each block. 7. The character area extracting unit merges the plurality of circumscribed rectangles when a plurality of the set circumscribed rectangles overlap each other or have a specific positional relationship in one of upper, lower, left, and right directions. 7. The specific object loading vehicle detection device according to claim 6, further comprising: means for extracting the character area from the rectangles merged by the merge means. 8. The character indicating the vehicle carrying the specific object is
The specific object loading vehicle detection device according to claim 5, wherein the device is "dangerous", "poisonous", or "dramatic".